Professional Documents
Culture Documents
1038/s41467-023-39236-4
Check for updates Deep cement mixing piles are a key technology for treating settlement distress
1234567890():,;
1234567890():,;
Subgrade settlement is a serious distress problem faced by the thereby improving the strength of the foundations12–16. It has the
construction of highways and railways1–7. It is particularly promi- characteristics of low specific gravity of consolidated body, large
nent in soft loess covered areas, river valley sections, mined out bearing capacity, low permeability coefficient, and has the advan-
areas, karst development areas, etc., which poses a huge threat to tages of low cost, short construction time and maximum utilization
the service life and operation safety of roads. Therefore, in the of the original soil. Therefore, it has been widely used in soft soil
process of subgrade construction, it is necessary to take effective ground improvement, foundation settlement treatment and other
measures by every means to improve the quality of subgrade con- engineering construction around the world17–24. At present, DCM
struction to prevent the occurrence of subgrade settlement piles have become one of the key technologies in highway distress
distress8–11. Deep cement mixing (DCM) pile is a kind of cement plus prevention and treatment engineering. However, some piles may
solid, which uses cement as the main curing agent and forcibly have different degrees of defects due to the dynamic conditions of
mixes the soft soil and cement paste in the deep part of the foun- soil layer and groundwater, construction technical defects, con-
dation through mixing machinery, making the soft soil harden into struction loopholes and other reasons. It has laid hidden dangers
a cement reinforced soil with integrity and certain strength, for inducing accidents25,26. Therefore, pile defect detection and pile
1
School of Earth Sciences and Engineering, Xi′an Shiyou University, Xi’an 710065, P.R. China. 2Shaanxi Key Laboratory of Petroleum Accumulation Geology,
Xi’an 710065, P.R. China. 3Shaanxi Geo-mining Geophysical and Geochemical Exploration Team Co. Ltd., Xi’an 710043, P.R. China. 4National Engineering
Research Center of Offshore Oil and Gas Exploration, Beijing 100028, P.R. China. 5Shaanxi Land Construction Surveying, Planning and Design Institute Co.
Ltd., Shaanxi Land Construction Group, Xi’an 710075, P.R. China. 6School of Highway, Chang′an University, Xi’an 710064, P.R. China. 7Gansu Luqiao Highway
Investment Co., Ltd., Lanzhou 730030, P.R. China. e-mail: shenhongyan@xsyu.edu.cn; 450991137@qq.com; Yanyueying@xsyu.edu.cn
construction quality evaluation are important guarantee technol- pile construction quality evaluation in other projects under high-
ogies for subgrade distress treatment. density pile groups, such as slope protection, dam seepage prevention,
The traditional engineering pile foundation detection mainly foundation reinforcement, and other engineering.
adopts static load testing, drilling coring, high-strain and low-strain
dynamic measurement. Static load testing27–29 and drilling coring30,31 Results
are relatively accurate and fair methods for pile foundation detection. Geological models and their GPR response characteristics
However, both of these methods are destructive testing techniques The purpose of using DCM piles in soft soil foundation is to enheance
with long cycles, high costs, low sampling ratios (1~2%) and poor the bearing capacity of the stratum and then to improve the stability of
representativeness of sampling. High-strain dynamic measurement32–34 the foundation. If the subgrade strengthened by DCM pile groups is
is to impact the pile top with a heavy hammer, and then analyze and regarded as a complete set of strata, the problem of pile foundation
determine the vertical compressive bearing capacity and pile integrity detection can be transformed into the problem of ground improve-
of a single pile based on the wave theory. This method can only provide ment quality evaluation (Fig. 1a, b). If there are quality problems in
reference bearing capacity. Low-strain dynamic measurement35,36 the foundation evaluation, it will also reflect that there are defects in
involves applying a dynamic force (dynamic load) to the pile top, and the pile foundations. The purpose of this is to create conditions for the
the structural integrity of the pile body is judged through observation selection of continuous detection technologies. It can significantly
and analysis of dynamic response signals. This method has the reduce the cost of pile foundation detection while effectively
advantages of fast detection speed, low cost and good reliability, and improving the efficiency and effect of pile foundation testing.
has a good detection effect for high-strength pile foundations, such as In order to effectively use GPR to detect the quality of pile foun-
cast-in-place piles, prefabricated piles, etc. Currently, pile foundation dations or evaluate the effectiveness of ground improvement, it is
testing is mainly based on this technology37–39. However, DCM piles are necessary to first understand the propagation laws of electromagnetic
locally sourced and mixed with cement paste to form piles. Although waves under different geological conditions of subgrade and the GPR
there are differences between the elasticity of the piles and the original response characteristics. For this reason, we analyzed the electro-
stratum, the difference is far less obvious than that of the piles cast magnetic waves propagation and GPR response characteristics of the
with sand, cement and coarse aggregates as the main material13,29. geological models of reinforced subgrade by pile groups. In the
Therefore, the low-strain dynamic measurement technology devel- foundations without pile (Fig. 1c), electromagnetic waves are reflected
oped on the basis of elastic theory may not be suitable for the detec- only at the wave impedance interfaces (stratum interfaces). For the
tion of DCM piles. In addition, the number of DCM pile groups is too foundations reinforced by complete piles (Fig. 1d), in addition to
huge, the pile spacing is small, and the pile group density is high for the forming reflections at the interfaces of wave impedances, electro-
soft soil reinforcement engineering. It is difficult to obtain obvious magnetic waves also generate complex reflection, transmission,
detection effect and high economic benefits by using this traditional refraction and guided waves between piles as well as between piles and
pile foundation testing technology. the original stratum. Thus, GPR response characteristics with relatively
Ground-penetrating radar (GPR) is a survey technology based on stable frequency and amplitude are formed in the equivalent forma-
electromagnetic field theory. Its basic principle is to transmit ultra- tion. For the foundations reinforced by defective piles (Fig. 1e), due to
high frequency (106~109 Hz) pulse electromagnetic waves, and then defects in the piles or stratum (such as pile breaking, necking, segre-
analyze and infer the structure and physical properties of under- gation, etc.), the energy distribution of electromagnetic waves at the
ground media according to the changes of echo amplitude, waveform, defect locations will be changed, which will also lead to change in GPR
frequency and other characteristics by using the differences of elec- response characteristics. Specifically, the amplitude become weaker
tromagnetic properties of underground media40–43. It has the advan- (type I response characteristics in Fig. 1e) or stronger (type II response
tages of high resolution, intuitive results and fast scanning speed. In characteristics in Fig. 1e), and the frequency also become lower or
recent years, it has become an important technology and means for higher.
various engineering quality detection, near- surface stratum structure In the quality evaluation process of ground improvement,
and defect survey44–55. Recent research has shown that there are sig- although the subgrade strengthened by piles is equivalent to a set of
nificant differences in electromagnetic properties between DCM pile strata, the GPR response characteristics of equivalent strata are com-
and original stratum, and there are also significant differences in pletely different from those of strata without pile, and there may be
electromagnetic properties between defective and complete piles56,57. significant differences between the GPR response characteristics of
Therefore, GPR technology is fully applicable to the quality inspection defective piles and complete piles (Fig. 1c, d, e). Therefore, these GPR
of DCM piles. response characteristics and differences become an important basis
Based on the above reasons, we proposed to use GPR technology for identifying whether piles have defects and evaluating the quality of
to detect the defects of DCM piles, and then realize the quality eva- ground improvement.
luation of ground improvement. Our innovations are to put forward
the idea of transforming defect detection of pile foundations into Technical system
quality evaluation of ground improvement, construct geological The outstanding characteristics of ground improvement by DCM piles
models of subgrade reinforced by pile groups and reveal theirs GPR are that the project progress is fast, the construction process is closely
response characteristics, develop a GPR multi-attribute information connected, and the time available for pile foundation defect detection
extraction technology, and establish a GPR technical system for eval- and ground improvement quality evaluation is limited. It is required
uating the quality of ground improvement. The pile length, strong/ that pile foundation defect detection and ground improvement quality
weak reinforcement stratum, primary stratum, the buried depth and evaluation not only to quickly and accurately detect the distribution
range of defective formation can be effectively identified by integrat- areas and scopes of defective piles or unfavorable stratum, but also to
ing the GPR information of single-channel waveform, multi-channel meet the needs of rapid construction of highway engineering, so as
section and attributes. Our research results provide a fast, efficient and to timely feedback on the detection results and take timely measures
economic way for the quality evaluation of DCM pile reinforcement for to supplement piles. Therefore, we have established a technical system
soft soil foundation, and fill the technical gap of efficient detection of for using GPR to detect defects in DCM piles and evaluate the quality of
pile foundation defects in the case of large number of pile groups with ground improvement based on engineering geological survey data and
small pile spacing and high pile group density. Moreover, our tech- highway construction site conditions (Fig. 2). Specifically, it includes
nology can also be expanded for pile foundation defect detection and the following four aspects.
Geological basis. Comprehensively master the stratum structure and electromagnetic wave velocities to achieve time-depth conversion
physical properties of subgrade as well as the thickness of soft soil processing of the GPR section.
layer based on engineering geological survey data. Deeply understand
design the parameters such as pile material, diameter, length and Project overview and GPR data acquisition
spacing as well as the technical scheme for subgrade construction. At present, DCM piles have been widely used in the treatment of
subgrade settlement distress caused by soft loess settlement24,58–60. We
Data acquisition. Design GPR survey lines or networks based on the have carried out the application and practice of GPR technology in
site conditions of highway construction. The basic principle is to lay defect detection of DCM piles and quality evaluation of ground
survey lines along the extension direction of the highway and along the improvement based on the construction project of Anlin highway in
axis of the pile foundations (Fig. 1b). For pile foundation detection with Gansu Province, China. Anlin highway stats from Anjiazui Village,
a length of several meters to more than ten meters, it is recommended Taishi Town, Linzhao County, and ends at Baichuan Village, Dongyuan
to select 100~200 MHz antennas and use profiling method to con- Township, Linxia County (Fig. 3a), with a total length of 56.7 km. It
tinuously collect GPR data. Note that the sampling length must com- would be constructed according to the technical standards of two-way
pletely include the equivalent layer depth of the reinforced subgrade four lane second-class highway.
by pile foundations, and the sampling rate is generally from 0.1 to It was found that there were signs of soft soil foundations in
0.4 ns to meet the detection requirements. Sanjiaji section during the highway survey stage. A combination of
drilling and excavation exploration was used to obtain foundation soil
Data processing. Carry out fine processing of GPR data around key samples. It was found by observing the foundation soil samples that
links such as energy enhancement in deep layers, interference noise the undisturbed soil particles in this section were mainly composed of
suppression and multiple attenuation. Ensure that GPR data proces- silt particles (particle size 0.006~0.078 mm), with obvious porosity.
sing results with high signal-to-noise ratio, high fidelity and high Macropores and wormholes were visible to the naked eye, and con-
resolution are obtained. tained a large mount of salt crystals. Therefore, it was determined that
this section belongs to typical soft loess. DCM pile technology was
Data interpretation. Integrate single-channel waveform, multi-channel selected to strengthen and treated the subgrade of this section in
section and attribute information, and combine drilling results to order to ensure high-quality construction of highway engineering. The
interpret geological information in GPR data. Realize the detection of design scheme for subgrade reinforcement is shown in Supplementary
pile foundation defects and accurate quality evaluation of ground Fig. S1. The DCM piles were arranged in a regular quincunx pattern,
improvement. It should be emphasized here that 1~2 boreholes can be with a pile diameter of 0.5 m, a pile spacing of 1.5 m, and a pile length of
properly arranged on the GPR survey line (or network), rather than 10.0 m. A total of more than 10,000 DCM piles were built in the entire
many boreholes. The purpose of borehole coring is not only to verify highway section.
the GPR interpretation results, but also to calibrate the GPR section 28 days after the completion of pile construction, the work of
with borehole results to further calculate and obtain more accurate using GPR technology to evaluate the quality of pile foundations and
a b Transmitting antenna
Receiving antenna
Ray path of direct waves
Ray paths of reflection at formation interfaces
Ray paths of electromagnetic waves about pile 1
Ray paths of electromagnetic waves about pile 2
Ray paths of electromagnetic waves about pile 3
Subgrade defect
GPR records
Pile
GPR GPR GPR I II
c d e
1 2 3 1 2 3
Fig. 1 | Geological models of ground improvement by pile groups and their GPR transformed into quality evaluation of ground improvement. c Stratum without
response characteristics. a Geological model of subgrade reinforced by pile pile, (d) stratum with complete piles, (e) stratum with defective piles and theirs GPR
groups. b Equivalent geological model of subgrade reinforced by pile groups. Pile response characteristics. Type I denotes energy weakening, and type II represents
groups are equivalent to a stratum, and defect detection of pile foundations is energy enhancement.
Fully understand subgrade structure, physical properties and thickness of soft soil layer as well
Geological Basis as engineering design parameters, such as pile foundation material, diameter, length and
spacing of piles.
Design GPR survey line or network according to the situation of the project site. Select
100~200 MHz antennas, and use the profiling method to collect GPR data. Pay attention to the
Data Acquisition
selection of data acquisition parameters, such as GPR record length, sampling rate and point
spacing.
Carry out fine processing of GPR data. Pay attention to gain recovery, energy compensation,
Data Processing
interference noise suppression, multiple attenuation and topographic correction.
Waveform attributes
Data Interpretation of GPR Amplitude attributes
Interpretation attributes Frequency attributes
Phase attributes
Fig. 2 | Technical system for DCM pile foundation defect detection and ground improvement quality evaluation via GPR. The technical system includes four parts
that are geological basis, data acquisition, data processing, and data interpretation.
a b c
China
Lanzhou
Yongjing
Dongxiang Anjiazui
Baichuan
Sanjiaji
Linxia
Hezhen Linzhao
Kangle
Fig. 3 | Location of work area and field data acquisition. a Location of work area. b DCM pile construction scene. c GPR data acquisition scenario.
the effect of subgrade reinforcement was carried out. We used the SIR- Interpretation of GPR section
4000® GPR system, selected 100 MHz antennas, and used the profiling Taking a survey line in the exploration area as an example, the appli-
method for GPR data acquisition. The scene of piling and field GPR cation effect of our proposed technology is introduced in detail.
data acquisition is shown in Fig. 3b, c. The data acquisition parameters Supplementary Fig. S2 shows the original GPR data obtained for this
included that the sampling rate was 0.351 ns, the sampling length was survey line and Fig. 4a shows the GPR data processing results
1024 points per channel, and the channel spacing was approxi- according to the data processing flow in Supplementary Fig. S3. Four
mately 2.0 cm. typical channels were extracted from the GPR section in Fig. 4a for
Depth (m)
Time (ns)
irregular circle was interpreted as a subgrade defect, the cyan solid line
was interpreted as a clear bottom interface of subgrade reinforcement,
200 10 and the cyan dotted lines were interpreted uncertain bottom inter-
faces of subgrade reinforcement.
Depth (m)
Time (ns)
Pile?
Pile? Pile bottom is difficult to give an accurate conclusion as to whether there are pile
Pile bottom? Pile bottom Pile bottom? foundations or defects in the left A and right B areas of the section.
200 10 Therefore, we used the GPR attribute information extraction and
analysis technology that we recently developed to further process the
C1 C2 C3 C4 section in Fig. 4a. It was attempted to clearly reveal whether the pile
300 15
foundation was complete and accurately evaluated the quality of weak
Distance (m)
c 0 10 20 30 40 50 subgrade reinforcement through GPR attribute information. Here, we
0 0
only selected one of the most representative GPR attributes from
waveform, frequency, amplitude and phase attributes categories for
in-depth analysis and geological information interpretation.
100 5
Depth (m)
Time (ns)
B
A Arc length of time window (AL). There must be differences in GPR
response characteristics between reinforced and primary stratum,
200 10
between strong and weak reinforced stratum as well as between
defective and non defective stratum. However, when the differences in
300 15 GPR response characteristics are not obvious, it is difficult to distin-
Bottom interface of piles (Determined) Defect boundary guish them in conventional GPR sections, such as the left A and right B
Bottom interface of piles (Uncertain) Borehole areas in Fig. 4c. It could be seen from the AL attribute in Fig. 5a that the
A, B Areas with unclear geological interpretations C1~C4 Labels of GPR channels boundary between strong and weak values clearly showed the bottom
interface of the strongly reinforced stratum (cyan solid line). The
Fig. 4 | Interpretation of GPR section. a GPR data processing results. b Waveform
interface was interpreted as a pile bottom interface in the GPR section
interpretation of typical channels. c Multi-channel section interpretation. Inter-
in Fig. 4c, which might be a wrong interpretation results. In addition,
pretation of GPR section is performed after analyzing and interpreting the wave-
the energy disturbance areas of this attribute section clearly deli-
form characteristics of typical channels. It is not possible to determine whether pile
foundations exist in A and B areas. Source data are provided as a Source Data file.
neated the defective areas of the subgrade (irregular green solid line
circles), which was clearer and more accurate than that in the GPR
section in Fig. 4a.
waveform feature analysis and interpretation in order to accurately
grasp the GPR response characteristics. The extracted GPR channels Product of instantaneous amplitude and cosine of instantaneous
were identified as C1, C2, C3 and C4, respectively (Fig. 4b). In the GPR phase (PIACIP). It could see from the PIACIP attribute in Fig. 5b that
channel marked with C1, the energy between 44 and 57 ns was weak the effect was equivalent to that of the AL attribute in Fig. 5a. The
and the waveform changed irregularly. It was interpreted that there bottom interface of the strongly reinforced stratum was clear (cyan
were subgrade (or pile foundation) defects between 44 and 57 ns of the solid line) as well as the defect areas of the subgrade were also clear
channel based on the analysis conclusion in Fig. 1e. The energy (irregular green solid line circles). It further confirmed that the inter-
between 122 ns and 163 ns of the channel was also weak, but a wavelet pretation of the AL attribute was reliable.
with strong energy appeared near 163 ns. Therefore, it was impossible
to judge whether the pile bottom was at 122 ns or 163 ns based solely Slope of instantaneous frequency (SIF). As could be seen from Fig. 5c
on the waveform characteristics. In the GPR channel labeled C2, there that the bottom interface of the strongly reinforced layer was not
was waveform disorder between 48 and 119 ns. This indicated that reflected in the section of SIF attribute, but it clearly revealed the
there were defects in the subgrade (or pile foundation). The GPR interface between the weakly reinforced layer and the primary stratum
energy suddenly weakened at 163 ns of this channel, which should be (bottom interface of the weakly strengthened layer, purple solid line).
the GPR response characteristic of the pile bottom. In the GPR channel The bottom interface information of the weakly reinforced layer could
labeled C3, the waveform changed were more regularly. It was judged not be obtained from the above two attributes in Fig. 5a, b and the
that the pile foundation was complete or there was no defect in the conventional GPR section in Fig. 4a. The disordered areas of this
subgrade based on the analysis conclusion in Fig. 1d. The amplitude attribute also clearly circled the defective areas of the subgrade (irre-
attenuation of the channel appeared at 131 ns, indicating that this gular green solid line circles).
might be the pile bottom. In the GPR channel marked with C4, there
was a weak and chaotic energy phenomenon between 97 and 159 ns, so Slope of reflection strength (SRS). We could see from Fig. 5d that the
it was impossible to determine whether the pile bottom was at 97 ns SRS attribute also accurately identified the boundary between the
or 159 ns. weakly reinforced layer and the primary stratum (purple solid line),
100 5 100 5
Depth (m)
Depth (m)
Time (ns)
Time (ns)
200 10 200 10
300 15 300 15
Distance (m) Distance (m)
c 0 10 20 30 40 50 d 0 10 20 30 40 50
0 0 0 0
100 5 100 5
Depth (m)
Depth (m)
Time (ns)
Time (ns)
200 10 200 10
300 15 300 15
Bottom interface of strongly reinforced stratum Bottom interface of weakly reinforced stratum Defect boundary Borehole
Fig. 5 | Interpretation of GPR attributes. a Arc length (AL) attribute with an strongly reinforced layer is interpreted through AL and PIACIP attributes. The
analysis time window of 1.0 ns. b Product of instantaneous amplitude and cosine of bottom interface of the weakly reinforced layer is determined through SIF and SRS
instantaneous phase (PIACIP) attribute. c Slope of instantaneous frequency (SIF) attributes, and the subgrade defect areas are delineated by combining four types of
attribute. d Slope of reflection strength (SRS) attribute. The bottom interface of the GPR attributes. Source data are provided as a Source Data file.
with an effect equivalent to the SIF amplitude in Fig. 5c. The disordered It can be seen from the comparison between the above borehole
areas of the attribute also clearly delineated the defective areas of the coring results and the GPR attribute interpretation results that the two
stratum (green irregular circles). are basically consistent. Based on the drilling results and GPR inter-
pretation results (Table 1), we further calculated the electromagnetic
Comparison between GPR results and borehole coring wave velocities. For the subgrade reinforced by DCM piles, there might
Based on the above interpretation of GPR data, we selected two be some differences in the physical properties of the medium due to
boreholes arranged near C2 and C4 in Fig. 4b for coring, which were the different construction quality at different locations, which might
marked as Borehole #1 (Fig. 6a, b) and Borehole #2 (Fig. 6c, d), lead to some deviations in the electromagnetic wave velocities
respectively. And the depth of Borehole #1 and Borehole #2 were between them. However, this difference was relatively weak according
approximately 11.0 m and 10.0 m, respectively. to the statistical results of Borehole #1 and Borehole #2 (Table 1).
The coring results for Borehole #1 were divided into five segments Therefore, we took the average value of velocities (vave ≈ 0.1 m · ns−1) as
(Fig. 6a, b). The first segment (0~2.37 m) and the third segment the final electromagnetic wave velocity. And all GPR data processing
(5.92~8.25 m) corresponded to the 0~47.3 ns and 119.5~163.5 ns seg- results in Figs. 4, 5 were subjected to time-depth conversion proces-
ments in the GPR attribute sections in Fig. 5, respectively. It was shown sing based on this average velocity.
as gray cement block with good cementation and complete pile body.
The second segment (2.37~5.92 m) corresponded to the 47.3~119.5 ns Evaluation of ground improvement
segment in the GPR attribute sections in Fig. 5. It was manifested as Based on the interpretation results of typical channels, multi-channel
broken cement blocks with large pores and poor integrity of the pile section and attributes of GPR data (Fig. 6e) as well as constrained with
body. This might be caused by the excessive acceleration of the nozzle boreholes (Fig. 6a, b, c, d), we had constructed a geological model
during construction, resulting in uneven mixing of cement and soil. (Fig. 6f) of the subgrade reinforced by DCM piles below this survey
The fourth segment (8.25~9.69 m) corresponded to the 163.5~193.7 ns line. The stratum after ground improvement could be divided into
segment in the GPR attribute sections in Fig. 5. It appeared as gray three layers that were strongly reinforced layer, weakly reinforced
cement blocks. The cementation property was not as good as that of layer and primary stratum. The average buried depth of the bottom
the first segment. And the integrity of the pile body was slightly poor. interface of the strongly reinforced layer was approximately 6.0 m.
The reason might be that the water content in the deep soil layer, The shallowest distributed on the right side of the section, approxi-
resulting in poor pile formation effect of DCM piles. The fifth segment mately 4.4 m. And the deepest was distributed in the middle left of the
(9.69~11.0 m) was primary loess. section, approximately 8.5 m. The average buried depth of the bottom
The coring results of Borehole #2 were divided into three segments interface of the weakly reinforced layer was approximately 9.0 m. The
(Fig. 6c, d). The first segment (0~4.91 m) corresponded to the 0~97.2 ns shallowest was approximately 7.2 m, distributed in the middle right of
segment in the GPR attribute sections in Fig. 5, and the core sample the section. And the deepest was approximately 10.0 m, distributed in
characteristics were the same as those of the first and third segments of the middle left of the section. There were two defect areas in the
Borehole #1. The second segment (4.91~8.26 m) corresponded to the stratum of the reinforced subgrade, which were identified as D1 and D2,
97.2~164.5 ns segment in the GPR attribute sections in Fig. 5, and the respectively. The defect area D1 presented an irregular shape and was
core sample characteristics were the same as the fourth segment of distributed on the right side of the section. It had a large area, spanning
Borehole #1. The third section (8.26~10.0 m) was primary loess. of approximately 24.0 m and a buried depth of 2.1~6.6 m. The defect
Distance (m)
a c e 0 10 20 30 40 50
Seriously defective 0 0
D2
D1
100 5
Depth (m)
Time (ns)
A B
200 10
Borehole #1 Borehole #2
b Borehole #1 d Borehole #2
0.00 0.00 300 15
Distance (m)
f 0 10 20 30 40 50
2.37 0
Depth (m) D2
D1
4.91 5
Depth (m)
Depth (m)
5.92
8.25 8.26 10
9.69 10.00
11.00 15
Determined pile bottom interface (by GPR waveforms) Uncertain pile bottom interface (by GPR waveforms)
Bottom interface of strongly reinforced stratum (by GPR waveforms) Defect boundary (by GPR waveforms)
Bottom interface of weakly reinforced stratum (by GPR attributes) Defect boundary (by GPR attributes)
Complete Slightly defective Seriously defective Primary loess Borehole
Strongly reinforced stratum Weakly reinforced stratum Defective area Primary stratum
A, B Areas with unclear geological interpretations (by GPR waveforms) D1, D2 Areas with geological defects
Fig. 6 | Construction of geological model after ground improvement. the GPR interpretation results constrained by dilled results. The reinforced sub-
a, b, c, d Drilling and coring results of Borehole #1 and Borehole #2. e Comparison grade is divided into three layers that are strongly reinforced stratum, weakly
of interpretation results of GPR waveforms (section) and attributes. f Geological reinforced stratum and primary stratum. There are two defects in the subgrade that
model after ground improvement. The geological model is constructed based on are D1 and D2. Source data are provided as a Source Data file.
Table 1 | Calculating electromagnetic wave propagation velocities by combining GPR interpretation and drilling results
Stratum interface Borehole #1 Borehole #2
Depth /m Double-way travel Velocity /m · ns−1 Depth /m Double-way travel Velocity /m · ns−1
time /ns time /ns
Top interface of D1 2.37 47.3 0.1002 - - -
Bottom interface of D1 5.92 119.5 0.0991 - - -
Bottom interface of strongly reinforced 8.25 163.5 0.1009 4.91 97.2 0.1010
stratum
Bottom interface of weakly reinforced 9.69 193.7 0.1001 8.26 164.5 0.1004
stratum
Average velocity /m · ns−1 - 0.1 - 0.1
area D2 presented an approximately flat elliptical shape and dis- construction process is carried out in accordance with standard pro-
tributed in the right side of the section. It spanned approximately cess as well as the construction information such as the improvement
13.0 m, had a thickness of approximately 1.0 m and a buried depth of length of each pile, pressure, speed, grout volume injected, etc., are
2.2~3.3 m. Therefore, it was recommended to supplement piles to the recorded, defects in pile body and inconsistent pile lengths can not be
defect D1 and D2 areas. avoided which creates potential safety hazards for highway engineer-
ing construction. This is why it is necessary to carry out pile foundation
Discussion testing.
DCM piles have played a key role in the treatment of foundation set- If there are defects in the pile foundations, the quality of ground
tlement distress17,19,21–24. However, piling is a complex and systematic improvement will inevitably decline. Therefore, there is a relationship
construction process. Small errors in each construction step may of mutual influence and restriction between pile foundation defects
cause deviations between the pile body and the design, such as and ground improvement quality. The ultimate purpose of pile foun-
insufficient (or uneven) mixing of cement paste and deep soil particles dation defect detection is to evaluate whether the quality of ground
as well as the lack of coordination between the lifting speed of the improvement has achieved good results. However, it is very challen-
grouting pipe and the grouting amount, etc. Even if the pile ging to accurately evaluate the quality of pile construction due to the
limitations of DCM pile material, large number of pile groups and small The signals that cause the GPR response will be very weak when the
pile spacing. If the single pile detection technology is adopted, and differences in the physical properties of underground media are quiet
then one pile after another is detected, not only will the detection weak. This makes it difficult to identify the information of formation
effect be poor, but also the detection efficiency will be low, and the media differences from weak GPR signals42. It can be seen from Fig. 6e
cost will be high. If the foundation strengthened by piles is regarded as that the traditional waveform (section) interpretation results omit to
a complete set of strata, the problem of pile foundation testing can be identify the defect area D2, and it is also difficult to accurately identify
transformed into a problem of ground improvement quality evalua- the bottom interface of the pile foundations. This is a manifestation of
tion. It can significantly reduce the detection cost while effectively inaccurate interpretation results due to insufficient utilization of
improving the detection efficiency. information carried by GPR data. Our research results show that the
Piling is actually a process of transforming the original formation, problem of fine quality evaluation of ground improvement can be
and the transformed formation medium will undergo significant effectively solved by integrating the GPR information of single-channel
changes14,59, which also will lead to changes in the electromagnetic waveform, multi-channel section and attributes. On the basis of pre-
properties of different underground locations. The degree of geolo- liminary evaluation results obtained from waveform (section) inter-
gical transformation varies, resulting in different changes in the pretation, comprehensive analysis and interpretation of multi-attribute
underground medium and varying electromagnetic characteristics55,56. information can not only accurately identify the depth of DCM piles and
This creates conditions for using GPR technology to detect the chan- subgrade defect areas, but also further subdivide the subgrade
ges in the stratum caused pile construction and the differences in the strengthened by pile groups into reinforced and weakly reinforced
changes in the strata media between different underground locations. stratum. It provides an important technical guarantee for the safe and
Difference in the electromagnetic properties of underground media efficient construction of soft soil subgrade reinforcement projects.
can cause changes in the GPR response, specifically manifested in The more data revealing a scientific or engineering problem is not
changes in the waveform, amplitude, frequency, phase and related necessarily the better, but whether it is typical and representative.
attributes of the GPR signals40,42,49,50,54,55. However, the propagation and Although we have only reported one GPR section, this GPR section has
response of electromagnetic wave fields under the condition of pile already included possible problems with DCM piles reinforcement of
groups are very complex, and it is not easy to fully reveal the propa- subgrade, such as deviation between pile length and design, defects in
gation laws of electromagnetic waves under such condition. For this pile foundations, and differences in the reinforcement quality of sub-
reason, we have constructed geological models of subgrade reinfor- grade in different sections. Therefore, our sample data fully possesses
cement by pile groups, and qualitatively discussed the electro- typical and representative characteristics. It should be emphasized
magnetic wave propagation phenomenon in this case. Compared with that our work is mainly to detect the defects of DCM piles and evaluate
the complex problem of pressure-controlled spherical cavity expan- the effectiveness of ground improvement based on GPR data. How-
sion in semi-infinite soil, although our models are simplified or idea- ever, it is not rule out that there are other scientific issues in ground
lized geological models, we believe that their GPR response improvement engineering that can be studied or explored through
mechanisms are identical. That is, GPR response is based on differ- GPR information. Multi-data samples may be an important way to
ences in the electromagnetic properties of media. As long as there are discover and solve unknown scientific issues.
differences in underground media, differences in GPR response char- Each technology will have more or less deficiencies (or limita-
acteristics will inevitably be appeared. Therefore, GPR technology is tions), and so does our technology. Specifically, if there are strong
fully applicable to the detection of geological defects caused by electromagnetic field sources (such as high-voltage lines, substations,
pressure-controlled spherical cavity expansion in semi-infinite soil. For wireless communication base stations, etc.) or large metal objects
the stationary oscillating waves appearing in the measured data (as (such as trucks, drilling machines, etc.) in the construction site envir-
shown in Fig. 4b), we speculate that this may be due to the guided wave onment, GPR technology will not achieve good detection results.
phenomenon generated by electromagnetic waves within or between Therefore, when using GPR survey, it is important to avoid these
piles. We believe that these waves have a positive significance for situations as much as possible. In addition, when the electromagnetic
judging the integrity of piles, and therefore can be considered as properties of the detection targets and the surrounding medium are
effective wave fields. close, it is easy to cause misjudgment, that is, GPR survey may not be
In theory, higher frequency and smaller spacing do contribute to applicable at all in this case. Moreover, GPR interpretation usually has
improving the resolution of detection. However, if the frequency of multiple solutions. The GPR response characteristics caused by dif-
GPR antennas is too high, the penetration depth of electromagnetic ferent media are close to each other or the same media causes dif-
waves will be decreased. According to the existing technical experi- ferent GPR response characteristics. What kind of interpretation can
ence of GPR40–55, the maximum effective detection depth of 100 MHz match the actual situation? The work experience of technicians will
antenna is approximately 50 m and the maximum effective detection play an important role in this case. Finally, it should be noted that our
depth of 250 MHz antenna is approximately 10 m in the survey of GPR technology can effectively distinguish the electromagnetic dif-
Quaternary strata. In addition, if the channel spacing is too small, it will ferences of different formation media, but it is difficult to achieve a
inevitably increase the workload, thereby increase survey cost. In our quantitative (or accurate) interpretation of the relationship between
case, we selected 100 MHz antennas for detection. The electro- the ultimate expansion pressure and the ground surface displacement
magnetic wave velocity v obtained by matching GPR section with the of the expansion pressure.
drilled core was approximately 0.1 m · ns−1 and the selected sampling In summary, the effectiveness of soft soil subgrade reinforcement
rate Δt was 0.351 ns. Therefore, the detection depth h of 1024 sampling depends entirely on the quality of pile construction, and defect
points per channel is approximately: detection of pile foundations can be regarded as an effective mon-
itoring measure for the quality of pile construction or as an insurance
h = v × 4t × n=2 = 0:1 × 0:351 × 1024=2 = 17:97m: ð1Þ against accidents33–39. Therefore, in order to truly improve the quality
of subgrade reinforcement by DCM piles, it is necessary to strictly
Our case shows that the data acquisition parameters, such as follow the specifications during the pile construction process, elim-
100 MHz antenna, 2.0 cm point spacing, 0.351 ns sampling rate, and inate all possible construction errors, and strictly control the con-
1024 sampling points per channel, are fully suitable for the quality struction quality. Cutting off the possibility of pile foundation defects
evaluation of soft soil subgrade reinforced by DCM piles with a length from the source is the eternal goal pursued by subgrade engineering
of approximately 10.0 m. construction.
indicate the rate of attenuation and absorption. 5. Shan, Y. et al. Iterative method for predicting uneven settlement
caused by high-speed train loads in transition-zone subgrade. J.
df Inc Transportation Res. Board 2607, 7–14 (2017).
SIF = , ð4Þ
dt 6. Guo, Y., Zhai, W. & Sun, Y. A mechanical model of vehicleslab track
coupled system with differential subgrade settlement. Struct. Eng.
where, fInc denotes instantaneous frequency, t forms travel time, and d Mech. 66, 15–25 (2018).
represents derivation operator. 7. Zhou, S. & Wang, B. Review of research on high-speed railway
subgrade settlement in soft soil area. Railw. Eng. Sci. 28,
Slope of reflection strength (SRS). The SRS attribute is defined as the 129–145 (2020).
change of instantaneous reflection intensity (i.e. instantaneous 8. Emadali, L., Motagh, M. & Haghighi, M. H. Characterizing post-
amplitude) with reflection time (Eq. 5). If the reflection intensity construction settlement of the Masjed-Soleyman embankment
remains throughout the analysis time window interval, the slope value dam, Southwest Iran, using TerraSAR-X SpotLight radar imagery.
will approach zero. If the reflection intensity gradually increases from Eng. Struct. 143, 261–273 (2017).
the top to the bottom of the time window, the slope value is positive, 9. Zou, S. et al. Rheological characteristics and one-dimensional iso-
on the contrary, the slope value is negative. The attribute can reflect tache modelling of marine soft clays. Mar. Georesources Geo-
the vertical distribution characteristics and interbedding conditions of technol. 37, 660–670 (2019).
the stratum (Fig. 5d). 10. Wang, Y., Li, J. & Sun, D. Jet-grouting in ground improvement and
rotary grouting pile installation: theoretical analysis. Geomech. Eng.
AIns ðt Þ 21, 279–288 (2020).
SRS = , ð5Þ
dt 11. Meng, X., Liao, H. & Zhang, J. Research on the collapsibility of loess
after water immersion. Nat. Hazards 109, 303–328 (2021).
where, AIns represents instantaneous amplitude, t forms travel time, 12. Cheang, W., Bergado, D. T., Suksawat, T., Jamsawang, P. & Vootti-
and d is derivation operator. pruex, P. Behavior and simulation of deep cement mixing (DCM)
and stiffened deep cement mixing (SDCM) piles under full scale
Calculation of velocity loading. Soils Found. 51, 307–320 (2011).
It is necessary to know the propagation velocity of electromagnetic 13. Taghavi, A., Muraleetharan, K. K. & Miller, G. A. Nonlinear seismic
waves when conducting time-depth conversion for GPR section. behavior of pile groups in cement-improved soft clay. Soil Dyn.
Through joint calibration of drilling results and GPR section, the buried Earthq. Eng. 99, 189–202 (2017).
depth of the formation can be determined, and then the double-way 14. Wonglert, A., Jongpradist, P., Jamsawang, P. & Larsson, S.
travel time of the formation interface can be picked up from the GPR Bearing capacity and failure behaviors of floating stiffened deep
section to calculated the electromagnetic wave propagation velocity, cement mixing columns under axial load. Soils Found. 58, 446–461
(2018).
2h 15. Zhou, M., Li, Z., Han, Y., Ni, P. & Wang, Y. Experimental study on the
v= , ð6Þ
t vertical bearing capacity of stiffened deep cement mixing piles. Int.
J. Geomech. 22, 04022043 (2022).
where, v represents the propagation velocity of electromagnetic 16. Zhu, S., Chen, C., Cai, H. & Mao, F. Analytical modeling for the load-
waves, h forms the depth of the formation interface, and t denotes the transfer behavior of stiffened deep cement mixing (SDCM) pile with
double-way travel time. rigid cap in layer soils. Comput. Geotech. 144, 104618 (2022).
17. Tan, T. S., Goh, T. L. & Yong, K. Y. Properties of Singapore marine
Data availability clays improved by cement mixing. Geotech. Test. J. 22,
All the data supporting the findings of this study are available within 422–433 (2002).
the main text and the Supplementary Information. The source data 18. Ignat, R., Baker, S., Larsson, S. & Liedberg, S. Two- and three-
generated in this study are provided in the Source Data file and also dimensional analyses of excavation support with rows of dry deep
available in Figshare under accession code https://doi.org/10.6084/ mixing columns. Comput. Geotech. 66, 16–30 (2015).
m9.figshare.22958399. Source data are provided with this paper. 19. Wu, Y., Diao, H., Jian, L. & Zeng, C. Field studies of a technique to
mitigate ground settlement of operating highways. J. Zhejiang Univ.
Code availability -Sci. A: Appl. Phys. Eng. 17, 565–576 (2016).
The source codes and respective compiled executable are provided as 20. Sun, Y. et al. Model test of the combined subgrade treatment by
Supplementary Software 1 file. The source code has also been made hydraulic sand fills and soil-cement mixing piles. Bull. Eng. Geol.
available in Figshare under accession code https://doi.org/10.6084/ Environ. 79, 2907–2918 (2020).
m9.figshare.22958399. 21. Moghadam, M. S., Zad, A., Yazdi, M. & Moghadam, M. J. Perfor-
mance of T-shaped and conventional cement-soil deep mixing piles
References to stabilize soft base of high-speed trains. Int. J. Geotech. Eng. 16,
1. Weng, X. & Zhang, L. Settlement prediction of soft subgrade under 208–222 (2022).
broaden embankment. J. Chang’ Univ. 31, 17–21 (2011). 22. Phutthananon, C. et al. Reliability-based settlement analysis of
2. Yang, J. & Feng, Q. A new method for measuring subgrade settle- embankments over soft soils reinforced with T-shaped deep
ment in high-speed railway by using a linear CCD. Measurement 46, cement mixing piles. Front. Struct. Civ. Eng. 16, 683–656 (2022).
1751–1756 (2013). 23. Shu, B. et al. Case study of solid waste based soft soil solidifying
3. Li, G., Nguyen, T. N. & Amenuvor, A. Settlement prediction of sur- materials applied in deep mixing pile. Buildings 12, 1193 (2022).
charge preloaded low embankment on soft ground subjected to 24. Jia, J., Zhao, Y., Jia, J., Xin, C. & Zhang, B. Research on the reinfor-
cyclic loading. Mar. Georesources Geotechnol. 34, 154–161 (2016). cement effect of cement-soil mixing pile on collapsible loess
4. Li, H., Zheng, X., Sheng, Y. & Ke, S. Differential settlements of foundation. J. Railw. Eng. Soc. 39, 18–24 (2022).
embankment treated by cement fly-ash gravel pile and sheet pile in 25. Wonglert, A. & Jongpradist, P. Impact of reinforced core on per-
freeway extension constructions. Int. J. Geomech. 17, formance and failure behavior of stiffened deep cement mixing
40170921 (2017). piles. Computers Geotech. 69, 93–104 (2015).
26. Fleming, B. J., Sritharan, S., Miller, G. A. & Muraleetharan, K. K. Full- 48. Schneiderwind, S., Mason, J., Wiatr, T., Papanikolaou, I. & Reicherter,
Scale seismic testing of piles in improved and unimproved soft clay. K. 3-D visualisation of palaeoseismic trench stratigraphy and trench
Earthq. Spectra. Prof. J. Earthq. Eng. Res. Inst. 32, 239–265 (2016). logging using terrestrial remote sensing and GPR: a multiparametric
27. Briaud, J. L., Ballouz, M. & Nasr, G. J. Static capacity prediction by interpretation. Solid Earth 7, 323–340 (2016).
dynamic methods for three bored piles. J. Geotech. Geoenviron. 49. Torbaghan, M. E. et al. Automated detection of cracks in roads using
Eng. 126, 1078–1084 (2001). ground penetrating radar. J. Appl. Geophys. 179, 104118 (2020).
28. Sakr, M. Comparison between high strain dynamic and static load 50. Pilecki, Z. et al. Identification of buried historical mineshaft using
tests of helical piles in cohesive soils. Soil Dyn. Earthq. Eng. 54, ground- penetrating radar. Eng. Geol. 294, 106400 (2021).
20–30 (2013). 51. Yamaguchi, T. & Mizutani, T. Detection and localization of manhole
29. Mostafa, H. H. & Elhakim, A. F. Static versus dynamic load tests on and joint covers in radar images by support vector machine and
piles socketed in Dubai limestone. Innovative Infrastruct. Solut. 7, Hough transform. Autom. Constr. 126, 103651 (2021).
298 (2022). 52. Bano, M., Tsend-Ayush, N., Schlupp, A. & Munkhuu, U. Ground-
30. Hou, Y., Mi, X. & Hong, X. The origin identification on defect cement penetrating radar imaging of near-surface deformation along the
mixing pile in the core sampling. Int. Conf. Electr. Technol. Civ Eng, Songino active fault in the Vicinity of Ulaanbaatar, Mongolia. Appl.
Lushan, China (2011). Sci. 11, 82429 (2021).
31. Gong, D. et al. Coring of antarctic subglacial sediments. J. Mar. Sci. 53. Li, C. et al. Layered subsurface in Utopia Basin of Mars revealed by
Eng. 7, 194 (2019). Zhurong rover radar. Nature 610, 308–321 (2022).
32. Wrana, B. Pile load capacity: Calculation method. Stud. Geo- 54. Noshahri, H., Van der Meijde, M. & Scholtenhuis, L. O. GPR surveys
technica et. Mechanica 37, 83–93 (2015). in enclosed underground sewer pipe space. Tunn. Undergr. Space
33. Alwalan, M. F. & El Naggar, M. H. Analytical models of impact Technol. Inc. Trenchless Technol. Res. 129, 104689 (2022).
force-time response generated from high strain dynamic load test 55. Tesic, K., Baricevic, A., Serdar, M. & Gucunski, N. Characterization of
on driven and helical piles. Computers Geotech. 128, 103834 ground penetrating radar signal during simulated corrosion of
(2020). concrete reinforcement. Autom. Constr. 143, 104548 (2022).
34. Tu, Y., El Naggar, M. H. & Wang, K. Nonlinear fictitious-soil pile 56. Dong, X., Song, Z., Zhang, S., Yang, B. & Wu, S. Application study of
model for pile high-strain dynamic analysis. Soil Dyn. Earthq. Eng. electrical resistivity characteristics of core samples from soil-
162, 107464 (2022). cement mixing piles. China Civ. Eng. J. 49, 88–94 (2016).
35. Dhutti, A. et al. Numerical modeling of ultrasonic guided wave 57. Si, F. et al. Experimental study on diameter detection of high
propagation and defect detection in offshore steel sheet piles. pressure jet grouting pile based on electromagnetic induction
Appl. Sci. 11, 4076 (2021). effect. Chin. J. Rock. Mech. Eng. 40, 3434–3442 (2021).
36. Guan, W. et al. Influence of the three-dimensional effect of pile-soil 58. Yang, Y., Liu, Y. & Ren, X. Experimental research on the factors
system on the vertical dynamic response of large-diameter piles in influencing strength of cement mixed saturated loess. J. Railw. Eng.
low-strain integrity testing. Energies 15, 9548 (2022). Soc. 33, 21–25 (2016).
37. Liao, S. T. & Roesset, J. M. Identification of defects in piles through 59. Zhou, Z., Wang, K., Feng, H., Tian, Y. & Zhu, S. Centrifugal model test
dynamic testing. Int. J. Numer. Anal. Methods Geomech. 21, of post-grouting pile group in loess area. Soil Dyn. Earthq. Eng. 151,
277–291 (1997). 106985 (2021).
38. Chow, Y. K., Phoon, K. K., Chow, W. F. & Wong, K. Y. Low strain 60. Zhou, M., Xing, L., Chen, Y. & Qiao, H. Settlement characteristics of
integrity testing of piles: three-dimensional effects. J. Geotech. split grouting pile composite foundation in loess areas. Adv. Civ.
Geoenviron. Eng. 129, 1057–1062 (2003). Eng. 2022, 8627124 (2022).
39. Wu, W. & Zhang, Y. A review of pile foundations in viscoelastic 61. Manataki, M., Vafidis, A. & Sarris, A. GPR data interpretation
medium: dynamic analysis and wave propagation modeling. Ener- approaches in archaeological prospection. Appl. Sci. 11, 7531 (2021).
gies 15, 9432 (2022). 62. Tronicke, J. & Böniger, U. GPR attribute analysis: there is more than
40. Davis, J. L. & Annan, A. P. Ground-penetrating radar for high- amplitudes. First Break 31, 103–108 (2013).
resolution mapping of soil and rock stratigraphy. Geophys. Pro- 63. Zhao, W., Forte, E., Pipan, M. & Tian, G. Ground Penetrating Radar
spect. 37, 531–551 (1989). (GPR) attribute analysis for archaeological prospection. J. Appl.
41. Bradford, J. H., Privette, J., Wilkins, D. & Ford, R. Reverse-time Geophys. 97, 107–117 (2013).
migration from rugged topography to image ground-penetrating 64. Jorge, L. P., Sauck, W. A. & Junior, A. O. S. GPR for mapping fractures
radar data in complex environments. Engineering 4, and as a guide for the extraction of ornamental granite from a
661–666 (2018). quarry: a case study from southern Brazil. J. Appl. Geophys. 58,
42. Shen, H. et al. Enhancement of GPR diffracted waves processing: a 177–187 (2006).
case study from Taiyuan, China. J. Environ. Eng. Geophys. 24, 65. Bradford, J. H. & Wu, Y. Instantaneous spectral analysis: time-
237–247 (2019). frequency mapping via wavelet matching with application to
43. Lombardi, F., Podd, F. & Solla, M. From its core to the niche: insights contaminated-site characterization by 3D GPR. Lead. Edge 26,
from GPR applications. Remote Sens. 14, 3033 (2022). 1018–1023 (2007).
44. Aranha, P. R. A., Augustin, C. H. R. R. & Sobreira, F. G. The use of GPR 66. dos Reis Júnior, J. A., de Castro, D. L., de Jesus, T. D. S. & Lima-Filho,
for characterizing underground weathered profiles in the sub- F. P. Characterization of collapsed paleocave systems using GPR
humid tropics. J. Appl. Geophys. 49, 195–210 (2002). attributes. J. Appl. Geophys. 103, 43–56 (2014).
45. Böniger, U. & Tronicke, J. Integrated data analysis at an archae- 67. Zhao, W., Forte, E., Colucci, R. R. & Pipan, M. High-resolution glacier
ological site: A case study using 3D GPR, magnetic, and high- imaging and characterization by means of GPR attribute analysis.
resolution topographic data. Geophysics 75, B169–B176 (2010). Geophys. J. Int. 206, 1366–1374 (2016).
46. Hebsur, A. V., Muniappan, N., Rao, E. P. & Venkatachalam, G. 68. Khwanmuang, W. & Udphuay, S. Ground-penetrating radar attri-
Application of ground penetrating radar for locating buried impe- bute analysis for visualization of subsurface archaeological struc-
diments to geotechnical exploration and piling. Int. J. Geotech. Eng. tures. Lead. Edge 31, 946–949 (2011).
7, 374–387 (2013). 69. Zhao, W., Forte, E. & Pipan, M. Texture attribute analysis of GPR data
47. Conyers, L. B. Analysis and interpretation of GPR datasets for inte- for archaeological prospection. Pure Appl. Geophys. 173,
grated archaeological mapping. Surf. Geophys. 13, 654–651 (2015). 2737–2751 (2016).